GoalUse what you learned about AI, advanced searching and data collected about citizens on new invented cases.
New words
machine learning · learning patterns from data instead of following hand-written rules for every casebias · an unfair leaning, which an AI can copy from its training datasecond use · data collected for one purpose being used for anotherexact phrase · words in quotation marks that must appear together in results
Explanation
AI means computers doing tasks that usually need human thinking. Machine learning finds patterns in many examples, which is why AI grew fast when data and computing power became cheap. AI can be fast and helpful, but it can also be wrong while sounding sure, and it can copy the unfairness in the data it learned from. When you need to find reliable information, a precise search helps: an exact phrase in quotation marks, a minus sign to leave a word out, a file type or a date. Meanwhile, cities and services collect data about citizens through cards, cameras and scanners. The key questions are why the data was collected, whether it is used for something else and who decides. Always check an important AI answer against a trusted source.
Idea
One question to ask
AI answer
How could I check it?
Training data
Who was left out?
Search
Which exact words or filter would narrow it?
Citizen data
Was it collected for this purpose?
Worked examples
Photo shows
Tool says
Problem
a snowy field
unknown
few winter photos in training data
A photo-sorting tool for a nature club learned from pictures taken mostly in summer. It labels winter scenes âunknownâ. It is copying a pattern from its examples, not understanding the picture.
Chatbot
The museum closes at 18:30 every day.
Jana
I will look at the museumâs own notice before I trust that.
A chatbot states a museumâs closing time with great confidence. The sensible response is to check the museumâs own notice.
Use of the visit list
Fits the purpose?
plan how many lifeguards are needed
yes
send visit times to an advertiser
no
A swimming-pool entry card records each visit. Using the counts to plan staff fits the purpose; passing visit times to a company that sells energy drinks does not.
Class activity
Mixed clinic (groups of three, paper). Your teacher sets three stations with invented cases. Station 1: an AI tool gives a wrong answer about a recipe in a confident voice. Write two questions you would ask, a way to check and what you would tell a younger pupil. Station 2: a townâs water meters were installed to plan pipe repairs, but the readings are later used to guess which families are away and send them holiday adverts. Say what the purpose was, what changed and who should decide. Station 3: write a search for a PDF guide on school composting that leaves out the word ârecipeâ, using at least two techniques. Rotate and compare answers with another group.
Practice
1
A pupil says: âThe picture in this post looks so real that the event must have happened.â Give two checks you would make before believing it.
Pictures can be edited or made by AI, so looks prove nothing. Check who posted it and whether they gain from it, look for the same event from two other trusted sources, and see whether the caption, date and place match.
2
Write a search to find PDF guides on school composting but not pages about recipes.
For example: "school composting" guide -recipe filetype:pdf. The quotation marks keep the words together, the minus sign removes recipe pages and the file type asks for PDFs. Symbols may differ between search engines.
3
Water meters were fitted to plan pipe repairs. Later the readings are used to guess when families are away. Name the original purpose and the change.
Data
Collected for
Later used for
water meter readings
repairs
guess when homes are empty
Original purpose: planning pipe repairs. Change: using the readings to learn about familiesâ movements, which is a new purpose that people did not agree to, so it breaks trust. Someone independent should decide, and people should be told.
4
Give one benefit and one risk of an AI tool that tells a school greenhouse when plants need water.
Benefit: it saves time and may use less water. Risk: it may be wrong, for example if its sensor is faulty, so a person should still look at the plants. It also depends on good data.